Cortical Responses to the 2F1-F2 Combination Tone Measured Indirectly Using Magnetoencephalography

Cortical responses to the 2f1-f2 combination tone measured indirectly using magnetoencephalography

͒ ͒ David W. Purcell,a Bernhard Ross ͑ß͒, Terence W. Picton, and Christo Pantevb Rotman Research Institute at Baycrest, 3560 Bathurst Street, Toronto, Ontario, M6A 2E1, Canada ͑Received 1 March 2007; revised 26 May 2007; accepted 29 May 2007͒

The simultaneous presentation of two tones with frequencies f1 and f2 causes the perception of several combination tones in addition to the original tones. The most prominent of these are at frequencies f2-f1 and 2f1-f2. This study measured human physiological responses to the 2f1-f2 combination tone at 500 Hz caused by tones of 750 and 1000 Hz with intensities of 65 and 55 dB SPL, respectively. Responses were measured from the cochlea using the distortion product otoacoustic emission ͑DPOAE͒, and from the auditory cortex using the 40-Hz steady-state magnetoencephalographic ͑MEG͒ response. The perceptual response was assessed by having the participant adjust a probe tone to cause maximal beating ͑“best-beats”͒ with the perceived combination tone. The cortical response to the combination tone was evaluated in two ways: ﬁrst by presenting a probe tone with a frequency of 460 Hz at the perceptual best-beats level, resulting in a 40-Hz response because of interaction with the combination tone at 500 Hz, and second by simultaneously presenting two f1 and f2 pairs that caused combination tones that would themselves ͑ ͒ beat at 40 Hz. The 2f1-f2 DPOAE in the external auditory canal had a level of 2.6 s.d. 12.1 dB SPL. The 40-Hz MEG response in the contralateral cortex had a magnitude of 0.39 ͑s.d. 0.1͒ nA m. The perceived level of the combination tone was 44.8 ͑s.d. 11.3͒ dB SPL. There were no signiﬁcant correlations between these measurements. These results indicate that physiological responses to the 2f1-f2 combination tone occur in the human auditory system all the way from the cochlea to the primary auditory cortex. The perceived magnitude of the combination tone is not determined by the measured physiological response at either the cochlea or the cortex. © 2007 Acoustical Society of America. ͓DOI: 10.1121/1.2751250͔ PACS number͑s͒: 43.64.Ri, 43.64.Jb, 43.64.Qh, 43.66.Ki ͓BLM͔ Pages: 992–1003

͒ I. INTRODUCTION Goldstein, 1966; Smoorenburg, 1972a, b . The 2f1-f2 nonlinearity is termed “essential” because it occurs at low stimulus The mammalian cochlea contains nonlinearities that levels ͑Goldstein, 1966͒. This is unlike a mainly linear sys- generate frequencies not present in the stimulus. The sim- tem distorting only at higher levels ͑von Helmholtz, 1877/ ͒ plest acoustic stimulus that can be used to study this phe- 1954; Goldstein, 1966 . The level and phase of the 2f1-f2 nomena is composed of two pure tones with frequencies f1 tone within the cochlea can be estimated by presenting an Ͼ and f2, where f2 f1. In response, the cochlea produces a set additional “cancellation” tone of the same frequency of combination tones, or distortion products, of which the ͑Zwicker, 1955; Goldstein, 1966, 1969͒. Perception of the cubic 2f1-f2 and the quadratic f2-f1 are the most prominent combination tone can be eliminated by carefully adjusting and most studied. the magnitude and phase of the cancellation tone until noth- With favorable stimulus conditions, combination tones ing is heard. Goldstein ͑1966͒ facilitated the subject’s ability are audible. The Italian violinist Tartini knew of their exis- to hear the cancellation effect by adding an additional probe tence as early as 1714 ͑Jones, 1935͒. As Jones ͑1935͒ re- tone slightly mistuned from the cancellation tone ͑e.g., ports, by 1801 it was known that the tone produced by a 4Hz͒, of sufficient level to beat with the combination tone. 32-ft pipe in a cathedral organ ͑about 16 Hz͒ could be gen- The cancellation tone was then adjusted to eliminate the erated in the ears of listeners by simultaneously activating a beating sensation by canceling the combination tone. Using 16-ft pipe ͑about 32 Hz͒ and an even shorter pipe producing the cancellation technique, the cochlear 2f1-f2 combination ͑ ͒ its fifth about 48 Hz through the 2f1-f2 combination tone, tone has an apparent level of an externally supplied tone that thus dispensing with the need for 32-ft pipes. is 20–40 dB below that of equal level primaries in the ex- ͑ The 2f1-f2 combination tone has been extensively stud- ternal ear canal where tone pair stimuli f2 and f1 have equal ied psychoacoustically ͑e.g., Zwicker, 1955; Plomp, 1965; levels L2 =L1; Goldstein, 1966; Smoorenburg, 1972b; Hall, 1975͒. Cancellation methods may overestimate the level of the combination tone for small f2 / f1 due to suppression ef- a͒ ͑ Author to whom correspondence should be addressed; Electronic mail: fects of the stimulus f1 Smoorenburg, 1972b; Shannon and [email protected]. Currently affiliated with the National Centre for Au- Houtgast, 1980; Zwicker, 1981͒. diology at the University of Western Ontario, London, Ontario, N6G 1H1, The psychoacoustically estimated levels of the combina- Canada. b͒ ͑ Currently affiliated with the Institute for Biomagnetism and Biosig- tion tones with respect to the stimuli i.e., 20–40 dB below nalanalysis at the University of Münster, Münster, Germany. the primaries͒ are similar to the levels of distortion products

992 J. Acoust. Soc. Am. 122 ͑2͒, August 20070001-4966/2007/122͑2͒/992/12/$23.00 © 2007 Acoustical Society of America measured in the mechanical motion of the basilar membrane is impossible to use time-domain measures such as the ͑Robles et al., 1997͒, the cochlear microphonic ͑Gibian and evoked N1 or sustained response. In contrast, steady-state Kim, 1982͒, receptor potentials of IHCs ͑Nuttall and Dolan, frequency-domain measurements of auditory evoked poten- 1990͒, and single fibers of the auditory nerve ͑Goldstein and tials and magnetic fields can separate the responses to the Kiang, 1968; Buunen and Rhode, 1978; Kim et al., 1980͒. stimuli and to the combination tones because they are at The response can be eliminated from recordings of nerve different frequencies. Unfortunately, steady-state responses fibers using cancellation techniques ͑Goldstein and Kiang, above 100 Hz derive mainly from brainstem sources ͑Herd- 1968; Goldstein et al., 1978͒. These results indicated that the man et al., 2002; Purcell et al., 2004͒, and combination tones 2f1-f2 combination tone is present in the cochlea, and sug- are most relevant at frequencies higher than 100 Hz. Steady- gested that it was processed at the same characteristic place state measurements cannot therefore directly measure the re- as an externally supplied tone of equal frequency. The 2f1-f2 sponse to the 2f1-f2 combination tone in cortex. To over- combination tone was later observed directly in the mechani- come these constraints, an approach based on beating used in cal motion of the basilar membrane at its characteristic place psychoacoustics was adopted ͑Goldstein, 1966; Furst et al., ͑Robles et al., 1991, 1997͒. 1988, unpublished͒.. Two pure tones that are similar in fre- The distortion product otoacoustic emission ͑DPOAE͒ at quency and magnitude will cause beating at the difference frequency 2f1-f2 is initially generated at the f2 characteristic frequency. Therefore, cortical correlates of the 2f1-f2 combi- place where the two stimulus tones interact most promi- nation tone could be measured indirectly by causing a 40-Hz nently ͑Talmadge et al., 1999; Shera and Guinan, 1999; beat either between a combination tone and an externally Knight and Kemp, 2000͒. Signals at the DPOAE frequency presented probe tone, or between two combination tones. then propagate on the basilar membrane both basally and The sensation of beating between a 2f1-f2 combination tone apically. The apical propagation will reach the characteristic and an externally supplied tone can be perceived relatively place for frequencies equal to 2f1-f2. The DPOAE measured easily by naïve participants with little training. The 40-Hz in the ear canal can be dominated by energy from either the steady-state magnetic fields generated in the auditory corti- ͑ f2 or the 2f1-f2 place depending on the stimulus parameters ces Mäkelä and Hari, 1987; Pantev et al., 1996; Gutschalk ͑Shera and Guinan, 1999; Knight and Kemp, 2000, 2001; et al., 1999; Schoonhoven et al., 2003͒ by these beats can be Dhar et al., 2005͒. Wilson ͑1980͒ found significant discrep- measured using MEG. ancies between the levels of the ear canal DPOAEs and psy- This indirect method of obtaining correlates of the choacoustic cancellation levels in the same subjects. These 2f1-f2 combination tone measures physiological responses to may have been in part due to changes in dominant DPOAE the difference-frequency type combination tone or “beat.” sources across stimulus conditions. However, Furst et al. Cochlear microphonic ͑Gibian and Kim, 1982; Cheatham ͑1988͒ and Zwicker and Harris ͑1990͒ also reported that and Dallos, 1997͒, IHC ͑Cheatham and Dallos, 1997͒, and DPOAE levels in the canal were low compared to those es- single neuron studies ͑Smoorenburg et al., 1976; Kim et al., timated psychoacoustically in the same subjects. Further- 1980͒ have shown that both cubic and quadratic combination more, the finding that psychoacoustic cancellation does not tones travel as waves on the basilar membrane from their eliminate the ear canal DPOAE highlights the fact that dif- initiation place where the stimuli interact at their own char- ferent processes are involved in the generation of the acteristic places. They are also present in nerve fiber record- DPOAE and the perception of the combination tone ͑Furst et ings both near the stimulus interaction site and at their own al., 1988; Zwicker and Harris, 1990͒. The net DPOAE signal characteristic places ͑Kim et al., 1980͒. Dolphin and Moun- reaching the canal is a function of the relative contributions tain ͑1993; Dolphin, 1997͒ have shown however that the of its two sources and the reverse transmission characteristics quadratic combination tone recorded from the scalp as an of the basilar membrane and middle ear. The perceptual re- auditory evoked potential is predominantly from the place of sponse to the combination tone likely relates to a different the interaction of the stimuli. Robles et al. ͑1997; comment mixture of the two sources. in Cheatham and Dallos, 1997͒ also found no evidence to Correlates of the 2f1-f2 combination tone have been support a basilar membrane origin of the quadratic combina- measured at most levels of the auditory system, including the tion tone. Half-wave rectification in the IHC and auditory ear canal ͑e.g., Kemp and Brown, 1984; Probst et al., 1991͒, nerve can produce the quadratic evoked potential ͑Lins et al., basilar membrane ͑Robles et al., 1991, 1997͒, cochlear mi- 1995͒. crophonic ͑Gibian and Kim, 1982͒, inner hair cell ͓͑IHC͒; The present paper reports two experiments: the first used ͔ Nuttall and Dolan, 1990; Cheatham and Dallos, 1997 , audi- an external tone to beat with a 2f1-f2 combination tone, and tory nerve ͑Kim et al., 1980͒, and auditory brainstem ͑Rick- the second studied the beating together of two separate ͒ man et al., 1991; Pandya and Krishnan, 2004 . The purpose 2f1-f2 combination tones. The beating frequency was set of this study was to determine whether correlates of the near 40 Hz since the cortical MEG response is easy to mea- ͑ ͒ 2f1-f2 combination tone could be measured in the human sure at this frequency Ross et al., 2000 . In both experi- ͑ ͒ auditory cortex using magnetoencephalography MEG . ments, the 2f1-f2 combination tone is probably generated The most serious difficulty in obtaining the cortical re- near the f2 region of the basilar membrane. The 2f1-f2 com- sponse to a 2f1-f2 combination tone is separation of the bination tone then propagates to its characteristic place 2f1-f2 response from the response to the stimulus tones that where it could interact in Experiment 1 with the externally generate it. Since the stimulus tones have 20–40 dB higher supplied probe tone, or in Experiment 2 with a second 2f1-f2 intensity, and are present simultaneously with the response, it combination tone of different frequency. However, interac-

J. Acoust. Soc. Am., Vol. 122, No. 2, August 2007 Purcell et al.: Magnetoencephalographic correlates of combination tones 993 tions between the combination tone and the probe tone ͑or between the two combination tones͒ could also take place at higher levels of the auditory pathway to produce the 40-Hz beat. The place of interaction was not the focus of this study. Rather the purpose was to observe the MEG correlates of the combination tone 2f1-f2 in the human auditory cortex.

II. METHODS A. Subjects Subjects reported normal hearing and were screened for thresholds below 20 dB HL at frequencies 500, 750, and 1000 Hz, which are in the range of stimuli used in this experiment. For the ﬁrst experiment ͑n=11, 2 males͒, the participants varied in age from 18 to 47 years. In the second experiment ͑n=2, 2 males͒, the ages were 24 and 33 years. The experiments were approved by the Research Ethics Board of the Baycrest Centre and written informed consent was obtained from each participant after the nature of the study was explained in accordance with the principles of the Declaration of Helsinki.

B. Experimental conditions

In Experiment 1, a stimulus tone pair of f2 =1000 Hz and f1 =750 Hz was used to elicit a combination tone of 500 Hz. An additional probe tone of similar frequency was used to produce beating with the combination tone. During the first part of the experiment, the participant adjusted the level of a probe tone at approximately 495.5 Hz to produce FIG. 1. Magnitude spectra of the stimuli used in the experiments. ͑A͒ The ͑ ͒ the strongest psychoacoustic perception of a slow beat at stimulus tone pair f1 and f2 and the probe tone fb used in Experiment 1. A 4.5 Hz ͑Yost, 2000, 167 pp.͒. This beating rate was much hypothetical combination tone within the auditory system is shown with the ͑ ͒ easier to hear than at rates near 40 Hz, where it is perceived dash line labeled fdp. B The two tone pairs used in Experiment 2. The higher frequency tone pair is marked with dash lines labeled f and f .The as “roughness” rather than “beats” ͑Fastl, 1977, 1990͒. This 1a 2a lower frequency tone pair is marked with solid lines labeled f1b and f2b. The level of probe tone was subsequently used in the MEG mea- leftmost two dashed lines represent hypothetical combination tones elicited surement, but the frequency was lowered to 460 Hz to pro- by the two tone pairs. The line labeled fdpa would be elicited by the higher duce a 40-Hz beat ͓see Fig. 1͑a͔͒. Frequencies near 40 Hz frequency tone pair, and the line labeled fdpb by the lower frequency tone pair. ͑C͒ The ECD stimulus tone pair used in both experiments to elicit elicit a larger and stable MEG response, and correspondingly robust responses for the purpose of dipole fitting. larger signal to noise ratio ͑SNR͒, than frequencies near 5 Hz ͑Ross et al., 2000͒. A significant 40-Hz response in the MEG of f2a=931 Hz and f1a=696 Hz, and f2b=626 Hz and f1b signal would show that the probe tone produced beats with =563 Hz ͓see Fig. 1͑b͔͒. These stimuli were designed to the 2f1-f2 combination tone. In pilot MEG measurements, a elicit two combination tones at frequencies f =460 Hz and 4-Hz beat was detectable in one participant, but the SNR was dpa fdpb=500 Hz. These two combination tones would beat at substantially less favorable than at 40 Hz. We therefore de- 40 Hz, provided their relative levels were similar. The beat- cided to employ the beat frequency that was best for each ing could then elicit an MEG response if the absolute levels ͑ measurement type 4.5 Hz for psychoacoustics and 40 Hz of the two combination tones were sufficiently high. for MEG͒. The psychoacoustic measurement of best-beats was used C. Stimulus generation to estimate the approximate magnitude of the 2f1-f2 combination tone so that the depth of modulation and correspond- 1. Psychoacoustic and DPOAE stimuli ing response could be maximized both perceptually ͑Zwis- Auditory stimuli for the psychoacoustic and DPOAE locki, 1953͒ and in the MEG record. After some brief measurements were generated with 16-bit precision at a rate training, the participants were able to perceive the beats and of 32 kHz using a National Instruments 6052E input/output to adjust the level of the probe tone to give a sensation of board. A Grason Stadler Model 16 audiometer was used to best-beats. The change in beat frequency between the psy- adjust the magnitude of the electrical stimulus signals f2 and ͑ ͒ choacoustic and MEG measurements from 4.5 to 40 Hz f1 prior to transduction into acoustic stimuli by a pair of should not have affected the modulation depth of the beating. Etymotic ER-2 transducers coupled to an Etymotic Experiment 2 also used beating to show the presence of ER-10B+OAE probe. Stimulus tone f2 was generated with a2f1-f2 combination tone, but did not use an external probe one channel, and f1 with the other to minimize acoustic dis- tone. Rather, the stimuli were two tone pairs with frequencies tortion within the stimulus system. The probe tone near the

994 J. Acoust. Soc. Am., Vol. 122, No. 2, August 2007 Purcell et al.: Magnetoencephalographic correlates of combination tones ͒ DPOAE frequency fdp was generated electrically with a bination tone with combination tone . In control trials, only Stanford Research Systems signal generator, and its level the stimulus tones f2a and f1a were presented, so no MEG was controlled by the participant using a Madsen Micro5 response at 40 Hz was expected. audiometer. The probe tone was produced acoustically using The reference trials, used in both experiments, were de- an Etymotic ER-3A transducer, and mixed acoustically with signed to produce a robust 40-Hz MEG response to deter- ͓͑ ͒ the f2 stimulus tone in a small coupler just prior to one of the mine an equivalent current dipole ECD , Ross et al., 2000, OAE probe’s inlet tubes. Tones were calibrated at every fre- 2002͔ in each hemisphere for the 40-Hz steady-state re- ͑ ͒ quency of interest with the OAE probe sealed in a Knowles sponses. As shown in Fig. 1 c , two tones of f2 =500 and ͑ DB-100 Zwislocki coupler Siegel, 1994; Whitehead et al. f1 =460 Hz were presented at the relatively high level of 1995͒. The stimuli were presented continuously to the left 70 dB SPL. These tones were the same frequencies as the ear at levels L2 =55 and L1 =65 dB SPL. combination and probe tones used in the test trials, but they were about 25 dB higher in level than the average level of 2. MEG stimuli the combination tones estimated psychoacoustically. This Auditory stimuli were generated with 16-bit precision at stimulus condition produced a relatively large 40-Hz mag- 11.025 kHz by a Sound Blaster AWE 64 card in a control netic field for the purpose of finding the equivalent cortical ͑ ͒ computer linked to the MEG acquisition system. Magnitude source a single ECD in each hemisphere. Henceforth this stimulus will be referred to as the ECD stimulus. of the electrical signals f2 and f1 were controlled with a Madsen OB822 audiometer prior to transduction to acoustic signals with a pair of Etymotic ER-2 transducers. These D. Recording and analysis transducers delivered the sound to a small coupler at the left 1. DPOAE ear through approximately 2.92 m of PVC tubing. The left Sound pressure in the ear canal was measured for ear was chosen for the experiment because the 40-Hz re- 61.44 s using the microphones integrated into the Etymotic sponse shows a right hemispheric and contralateral domi- acoustic probe. The microphone signal was amplified using nance ͑Ross et al., 2005͒. the +40-dB setting on the supplied Etymotic amplifier, and In the first experiment, f was generated on its own 2 bandpass filtered 40–4000 Hz using a Krohn-Hite Model channel, whereas the f channel also produced the probe tone 1 3322 filter with a 24-dB/octave slope. The conditioned mi- during test trials. For the second experiment, the higher tones crophone signal was digitized at 16 kHz with 16-bit preci- of each pair ͑f and f ͒ were produced on one channel and 2a 2b sion. A discrete Fourier transform ͑DFT͒ was performed on the lower tones on another ͑f and f ͒. Signals were cali- 1a 1b the 61.44-s window, and the frequency bin containing f brated using the Zwislocki coupler, and no significant distor- dp was evaluated to determine whether the emission magnitude tions were present at the response frequencies. was significantly larger than the noise in six frequency bins In both MEG experiments tones were presented to the on either side using an F-test with 2 and 24 degrees of free- left ear for trials of duration 2.3 s. Presenting limited- dom ͑John and Picton, 2000͒. duration trials during MEG measurements allowed intermin- gling of control and test trials. This was intended to minimize time and motion confounds from the MEG data within 2. MEG recording and analysis individuals. These trials were sufficiently long to obtain a The MEG was measured by means of a whole head steady-state response and to perform spectral analysis. device with 151 sensors ͑Omega-151a, VSM Medtech͒ Both MEG experiments had three stimulus conditions: sampled at 625 Hz with a third-order gradient noise reduc- test trials, control trials, and reference trials. The test trials tion. Trials with muscle artifacts on any channel due to in Experiment 1 used the acoustic stimulus shown in Fig. events such as eye or jaw movements ͑approximately 20% of ͑ ͒ ͒ 1 a , where the stimuli, f2 and f1, as well as the probe tone fb total were removed from further analysis. The signal in each ͑ were presented the dotted line labeled fdp is an example of a trial was referred to a baseline value equal to the mean in a potential combination tone generated in the cochlea͒. Control 50-ms window immediately prior to the stimulus onset. For trials included the stimuli f2 and f1, but no probe tone. The each participant and stimulus condition, trials were averaged 40-Hz beat was expected in the MEG test trial measurements together and the average trials were bandpass filtered within since the 2f1-f2 combination tone should generate beats with the range 20–60 Hz. the probe tone. This response will be referred to as probe The estimated ECDs were localized in a Cartesian coor- beating ͑combination tone with probe tone͒. In the control dinate system with the y axis between the left and right ex- trials, there was no signal present to beat at 40 Hz with the ternal ear canals, the x axis from the center of this axis to the 2f1-f2 combination tone. nasion, and the vertical z axis orthogonal to the others. A In Experiment 2, the test trial stimulus included two tone spherical head model of radius 7.5 cm with the head center ͑ ͒ pairs shown in Fig. 1 b , where stimulus tones f2a and f1a at coordinates x=0, y=0, and z=5 cm was employed to de- were expected to produce the combination tone fdpa in the termine the locations of the ECDs. ͓ cochlea. Similarly, the stimulus tones f2b and f1b produce the The average trial elicited with the ECD stimulus Fig. ͑ ͔͒ combination tone fdpb. During test trials, if the combination 1 c was subsequently fit with a model sinusoid of 40 Hz on tones fdpa and fdpb were of similar magnitude, they could beat all channels in each hemisphere. The field of a single dipole with one another to generate a 40-Hz response in the MEG was then fit in each hemisphere for this 40-Hz model data record. This response will be referred to as CT beating ͑com- using the spherical head model. Although in most partici-

J. Acoust. Soc. Am., Vol. 122, No. 2, August 2007 Purcell et al.: Magnetoencephalographic correlates of combination tones 995 pants the dipole in each hemisphere was independently lo- TABLE I. Mean right and left hemisphere dipole sources determined from cated, for two individuals the smaller dipole in the left hemi- the MEG stimulus designed to elicit a robust 40-Hz steady-state response where tones were f2 =500 Hz and f1 =460 Hz presented at 70 dB SPL. Data sphere had to be fixed symmetrical to the one in the right are from ͑eight͒ participants in experiment 1 and the two individuals from hemisphere to avoid an unrealistic solution. Both dipoles Experiment 2. Dipole source coordinates are with respect to the origin of the were then fixed in position and orientation, and assumed to spherical head model. In this model the center of the head was presumed at be the location of the sources of any 40-Hz steady-state re- x=0, y=0, z=5 cm with sphere radius 7.5 cm. The x-z plane is sagittal with sponses in the test and control conditions. positive x toward the nose and positive z upwards. The y-z plane is coronal with positive y toward the individual’s left ear. The x-y plane is axial. Average trial data from the test and control conditions Standard deviation given in parentheses. ͓elicited with the stimuli in Fig. 1͑a͒ for Experiment 1, and Fig. 1͑b͒ for Experiment 2͔ were used to determine current Left hemisphere Right hemisphere dipole moments over time at the single dipole fixed in each 40-Hz steady-state dipole 40-Hz steady-state dipole hemisphere using the method of source space projection x ͑cm͒ 1.4 ͑0.9͒ 2.0 ͑0.8͒ ͑ ͒ Ross et al., 2000, 2002 . These dipole moments were then y ͑cm͒ 4.1 ͑1.0͒ −3.8 ͑0.5͒ evaluated for the presence of a 40-Hz beat in the frequency z ͑cm͒ 7.0 ͑0.7͒ 6.9 ͑0.6͒ domain using a DFT. Spectral analysis was performed from Magnitude Q ͑nA m͒ 1.5 ͑1.2͒ 2.3 ͑1.2͒ 250-ms poststimulus onset to the end of the trial such that there were an integer number of cycles of 40 Hz in the analysis window ͑total length 2.05 s͒. The initial part of the titled movie during the 1-h measurement session and were source wave form was excluded because it takes about instructed to remain as still as possible during the presenta- 250 ms for the steady-state response to become well estab- tion of tones. During each of five blocks lasting 10 min, 230 lished ͑Ross et al., 2002͒. Signal magnitude at 40 Hz was randomly intermingled control and test trials were collected, compared, as for the DPOAE, to noise in six bins above and each of 2.3 s with 150-ms interstimulus intervals. Between six below 40 Hz using an F-test with 2 and 24 degrees of blocks the subject could relax, but was instructed to do their freedom. best not to change their head position for the course of the experiment. The stimulus tone pair from the first part of the E. Experimental protocol experiment ͑same frequencies and levels͒ was presented to 1. Psychoacoustic and DPOAE the left ear for all trials. For test trials, the 460-Hz probe tone was also presented at the best-beat level. A final block of 230 During the first part of Experiment 1 when psychoacous- trials was collected with the ECD stimulus at 70 dB SPL for tic and DPOAE measurements were performed, participants the purpose of dipole fitting. Head localization was per- ͑ ͒ were seated in an Industrial Acoustics Company IAC sound formed at the start and end of each block of trials. insulated room. An acoustic probe was sealed in the left ear Experiment 2 was similar to Experiment 1 except that canal using an appropriately sized rubber tip. Participants the test trials used two tone pairs rather than a tone pair and were familiarized with controlling the level of the 495.5-Hz a probe. The first five blocks contained randomized control probe tone using an audiometer, and the sensation character- and test trials, where either the upper tone pair ͑control͒,or istics of a 4.5-Hz beat were demonstrated using example both tone pairs were presented ͑test͒. The tones were each tones. During the actual psychoacoustic adjustment, stimulus presented at 65 dB SPL. The same final block used the ECD tones f2 and f1 were presented to the left ear at L2 =55 and stimulus again at 70 dB SPL, as in Experiment 1. L1 =65 dB SPL, respectively. These relative levels were chosen to maximize the magnitude of DPOAEs ͑Gaskill and Brown, 1990͒. Each participant started with a low probe tone III. RESULTS level, and incremented it in 5-dB steps until the level for A. Experiment 1: Beats between combination tone best-beats had been surpassed. The participant then lowered and probe tone the probe tone level and fine-tuned it in 1-dB steps to maxi- One of the 11 participants tested did not show a signifi- mize the sensation of beating. Finally, the participant inter- cant 40-Hz beat response in the test condition of the MEG rupted the probe tone to demonstrate that the beating sensa- measurement for either hemisphere. This person did not have tion disappeared in the absence of the probe. The best-beats atypical DPOAE magnitude ͑−6.9 dB SPL͒, or did he select level of the probe tone was recorded for subsequent use in an unusual best-beat probe tone level ͑42.7 dB SPL͒.Inre- the MEG protocol. After the psychoacoustic assessment was sponse to the ECD stimulus, he had dipole moment magni- finished, physiological measurements of the individual’s tudes larger than the average values in Table I, but well DPOAE at 500 Hz were made for 61.44 s each, alone and within the range found for other participants ͑left hemisphere with probe tones of either 495.5 and 460 Hz. Q=2.6 nA m, right hemisphere Q=2.8 nA m͒. However, MEG noise levels during the test condition were the highest 2. MEG of any participant ͑beyond the mean plus 2 s.d. of the noise The second part of Experiment 1 took place in the mag- levels estimated for the others͒, and this was likely the rea- netically shielded room of the MEG. The subject sat upright son for his lack of a detectable MEG response. He was ex- with their head touching the top of the dewar’s cavity. Head cluded from further analysis. localization coils were located on inserts placed in both ears, With two other participants, the experimental protocol as well as at the nasion. Participants watched a silent sub- had to be altered to suit their needs. In one person, the best-

996 J. Acoust. Soc. Am., Vol. 122, No. 2, August 2007 Purcell et al.: Magnetoencephalographic correlates of combination tones TABLE II. Mean responses across eight participants from individual average test and control trials in Experi- ment 1. The three entries for DPOAE magnitude are without probe tone, with 495.5-Hz probe, and with 460-Hz probe, respectively. The three entries for best-beat probe level are as measured in the Zwislocki coupler with the 495.5-Hz probe tone, as measured in the ear canal with the 495.5-Hz probe, and as measured in the ear canal with the 460-Hz probe, respectively.

Sound booth MEG left hemisphere MEG right hemisphere

Source space source space Probe tone DPOAE Best-beat projection Noise projection Noise frequency magnitude probe level magnitude estimate magnitude estimate ͑Hz͒ ͑dB SPL͒ ͑dB SPL͒ ͑nA m͒ ͑nA m͒ ͑nA m͒ ͑nA m͒

Test trial No probe 2.6 ͑12.1͒ 44.8 ͑11.3͒ 0.28 ͑0.14͒ 0.09 ͑0.03͒ 0.39 ͑0.10͒ 0.07 ͑0.02͒ mean ͑s.d.͒ 495.5 1.1 ͑12.3͒ 46.2 ͑12.8͒ 460 1.7 ͑12.1͒ 44.7 ͑13.0͒ ͑ ͒ ͑ ͒ ͑ ͒ ͑ ͒ Control trial ¯¯¯0.05 0.03 0.07 0.03 0.07 0.04 0.06 0.01 mean ͑s.d.͒ beat probe tone level of 74 dB SPL was too high to be used individual was near significant at pϽ0.09. As shown in the with the MEG stimulus presentation system. A second person third and fourth columns of Table II, there was substantial objected to the loudness of the tones in the MEG measure- variability in both the DPOAE and the level of the probe ment, despite the same level having been used to measure tone chosen to generate best-beats. The probe tone level was DPOAEs. For both of these individuals, the MEG measure- significantly larger than the measured DPOAE level using a ment proceeded with lower stimulus levels ͑10 and 5 dB two-sample separate variance t-test ͑pϽ0.001͒. DPOAEs lower, respectively͒. Significant 40-Hz beat responses were were also measured with the probe tones of 495.5 and found for both individuals during the test condition ͑only the 460 Hz present. All three DPOAE measurements are pre- right hemisphere for the first, and both hemispheres for the sented in Table II ͑no probe, 495.5-Hz probe, and 460-Hz second person͒. However, since the protocol was changed for probe͒. Although the means are slightly lower, paired t-tests these individuals, they were also excluded from further showed no significant changes in DPOAE magnitude with a analysis. probe tone present. Figure 2 plots the ear canal spectra from All eight of the remaining participants in Experiment 1 a single individual for all three probe conditions. Although had right hemisphere source space projection dipole moment magnitudes, elicited by the 40-Hz probe beat during the test condition, that were statistically larger ͑using the F-test͒ than the noise at nearby frequencies. Seven of the eight participants also had statistically significant sources in the left hemisphere for the test condition. The dipole sources in the left and right hemispheres that were fit to the ECD stimulus ͑and then fixed in position and orientation͒ are shown in Table I. These dipoles were used to estimate the source space projection dipole moment magnitudes reported in Table II. The data shown in Table I are the average of the eight individuals who had significant responses in the test condition and followed the protocol of Experiment 1, and the two individuals in Experiment 2 since the same ECD stimulus was used to estimate these dipoles. Using paired t-tests, there were significant differences between some of the dipole parameters for the left and right hemispheres. For the y-axis parameter, the absolute value was used to compare the two hemispheres. The right source was significantly ͑pϽ0.01͒ more anterior than the left. Most strikingly, the response was much larger in the right hemisphere ͑pϽ0.0001͒. Every participant had a larger dipole moment magnitude in the right hemisphere compared to his or her own left hemisphere. Table II shows the DPOAE, psychoacoustic, and MEG data from the eight participants who had significant MEG FIG. 2. Magnitude spectra of the ear canal sound pressure from a single responses in the right hemisphere for test trials ͑pϽ0.001͒ in individual in Experiment 1. The three panels show data from the three probe Experiment 1, and followed the same standard experimental conditions where there was no probe tone, the probe was 4.5 Hz below the DPOAE frequency, and the probe tone was 40 Hz below the DPOAE fre- protocol. All but one of these participants also had significant quency, respectively. The noise floor fluctuates between the measurements, responses in the left hemisphere at pϽ0.05; the remaining but the DPOAE magnitude is unchanged.

J. Acoust. Soc. Am., Vol. 122, No. 2, August 2007 Purcell et al.: Magnetoencephalographic correlates of combination tones 997 FIG. 3. Magnitude spectra of the source space projection dipole moments FIG. 4. Magnitude spectra of the source space projection dipole moments from each hemisphere calculated from the average of the test trials ͑solid from each hemisphere calculated from the average of the test trials ͑solid line͒ and control trials ͑dash line͒ for the individual from Fig. 2. Note the line͒ and control trials ͑dashed line͒ for the individual from Experiment 2 spurious peak in the spectrum of the average control trial in the left hemi- with a significant response. sphere at 39 Hz. The left hemisphere response at 40 Hz in the average test trial only achieved a significance level of pϽ0.04. similar to those for the eight subjects used in Experiment 1 the noise floor fluctuates from one recording to the next, the and are included in the average data of Table I. In the test stimuli and DPOAE do not change levels between panels. condition, one participant had a robust 40-Hz CT beat re- The last four columns in Table II show the MEG source sponse in the right hemisphere ͓Q=0.245 nA m, noise esti- space projection dipole moment magnitudes and noise estimate 0.053 ͑s.d.=0.026͒ nA m, pϽ0.0001͔, and a small re- mates for the left and right hemispheres. Using paired t-tests sponse in the left hemisphere ͓Q=0.106 nA m, noise for both hemispheres, the source space projection dipole mo- estimate 0.043 ͑s.d.=0.022͒ nA m, pϽ0.02͔. The right ment magnitude was significantly larger in the test condition hemisphere response in the test condition showed a clear than in the control ͑no probe͒ condition ͑left pϽ0.002; right spike relative to noise in nearby frequencies as shown in Fig. pϽ0.0001͒. Although the mean source space projection di- 4. For this individual during the control condition, there was pole moment magnitude for the right hemisphere in the test no significant response for the left hemisphere ͓Q condition is larger than for the left, the difference was not =0.058 nA m, noise estimate 0.047 ͑s.d.=0.022͒ nA m, p quite statistically significant ͑paired t-test pϽ0.07; sign test Ͻ0.3͔ and a borderline response in the right hemisphere ͓ pϽ0.07͒. The right source space projection dipole moment Q=0.105 nA m, noise estimate 0.048 ͑s.d.=0.027͒ nA m, p magnitude was larger in all participants except one. The Ͻ0.04͔ that was likely a type I statistical error. The second probe beat responses are shown in Fig. 3 for the same indi- participant did not show any significant responses in either vidual whose DPOAE data are plotted in Fig. 2. The 40-Hz test or control conditions. Both participants had clear re- response in the left hemisphere ͑ipsilateral to the left stimu- sponses to the ECD stimulus ͑listed for left and right hemi- lus ear͒ was half the magnitude of the signal in the right spheres: 0.84 and 1.89 nA m for the participant with a CT hemisphere, although both were statistically different from beat, and 0.92 and 1.67 nA m for the one without͒. the background noise ͑pϽ0.04 and pϽ0.00001, respec- tively͒. IV. DISCUSSION Pearson’s r coefficients were calculated between measurement types across the eight participants from Experiment All participants in Experiment 1 had measurable 2f1-f2 1. There were no significant correlations between DPOAE DPOAEs that were statistically different from the back- magnitude, best-beat probe tone level, or MEG source space ground noise in the ear canal when the presented frequencies projection magnitudes in either hemisphere. were 750 and 1000 Hz, and the corresponding DPOAE was 500 Hz. All individuals were able to perceive beating at B. Experiment 2: Beats between two combination 4.5 Hz between an externally supplied probe tone and the tones 2f1-f2 combination tone and were able to adjust the level of The 40-Hz steady-state dipoles determined from the the probe tone to maximize this sensation and achieve best- ECD stimulus for the two participants in Experiment 2 were beats. When the frequency of the probe tone was changed so

998 J. Acoust. Soc. Am., Vol. 122, No. 2, August 2007 Purcell et al.: Magnetoencephalographic correlates of combination tones TABLE III. Comparisons between average DPOAE magnitude and standard deviation ͑s.d.͒ obtained in the present study, and from the literature using similar stimulus parameters ͑N/A denotes value not explicitly given͒.

͑ ͒ f2 L1 L1 −L2 Participants Ldp s.d. ͑ ͒ ͑ ͒ ͑ ͒ ͑ ͒ ͑ ͒ Study Hz f2 / f1 dB SPL dB ears dB SPL

Present 1000 1.33 65 10 8 ͑8͒ 2.6 ͑12.1͒ Harris et al. 2000 1.33 65 0 5 ͑10͒ −2 ͑6͒ ͑1998; Fig. 2͒ Abdala 1500 1.353 60 10 10 ͑11͒ −3 ͑6͒ ͑1996; Fig. 5a͒ Brown et al. 1700 1.3 60 15 40 ͑40͒ −7 ͑N/A͒ ͑2000; Fig. 2͒ Gorga et al. 750 1.2 65 15 80 ͑80͒ 0 ͑8͒ ͑1993; Fig. 3͒ Vinck et al. 830 1.213 60 0 101 ͑101͒ 3.6 ͑6͒ ͑1996; Fig. 1, Table 4͒

that the beating was at 40 Hz, all but one individual had below the level of the f1 stimulus tone, which is in the range robust 40-Hz MEG responses in the right hemisphere con- of that reported previously ͑Goldstein, 1966; Smoorenburg, tralateral to the stimulus. Most participants also had statisti- 1972b; Hall, 1975͒. The best-beat level of the probe tone was cally significant responses in the left hemisphere. Experi- more than 40 dB higher than the ear canal measurement of ment 2 looked at whether 40-Hz beating responses could be the DPOAE at frequency fdp=2f1-f2. This is consistent with obtained between two combination tones, each produced the within-subject results of Zwicker and Harris ͑1990͒ who from a pair of externally presented stimuli. One of two par- found the cancellation tone for the perceptual combination ticipants showed a 40-Hz MEG response with good SNR in tone to be 33–60 dB higher than for the ear canal DPOAE, the right hemisphere. The 40-Hz steady-state MEG signals depending on the individual and stimulus conditions. recorded in these two experiments objectively demonstrate The best-beat probe tone is similar in frequency to fdp, activation of the human primary auditory cortex in response and would itself excite the basilar membrane near the fdp to 2f1-f2 combination tones. characteristic place. To generate a beating sensation, the probe tone amplitude must be similar to the amplitude of the combination tone somewhere in the system in order to A. Psychoacoustic and DPOAE response achieve the most prominently perceived amplitude modula- The absolute level of the average DPOAE recorded in tion. Depth of modulation thresholds, from temporal modu- the ear canal ͑2.6 dB SPL͒ is typical for human subjects at lation transfer functions, suggest that modulation would be wide stimulus frequency ratios and low emission frequencies impossible to detect if the probe and combination tones dif- ͑as summarized in Table III͒. The s.d. reported here ͑12 dB͒ fered by more than 30 dB ͑Kohlrausch et al., 2000; Strick- is larger than previously reported values ͑6–8 dB͒. This may land, 2000͒. Therefore, it is unlikely that the DPOAE as be in part because we were able to include individuals with measured in the ear canal was causing the perceived beat. low-level emissions due to the long recording time ͑61 s͒. At least two possibilities could explain the discrepancy Such individuals may be excluded from studies with short, in levels between the probe tone used to estimate the per- clinically oriented measurements due to higher noise floors. ceived combination tone and the recorded DPOAE. In the At the stimulus frequencies used in these experiments first, the reverse transmission of the fdp signal on the basilar ͑ ͒ ratio f2 / f1 =1.33 , it is likely that a single dominant DPOAE membrane and across the middle ear may have significantly source is at the f2 characteristic place where the stimuli in- attenuated the energy of the DPOAE. The DPOAE emitted teract most prevalently. A second source at the 2f1-f2 char- into the canal could then under-represent the size of the acteristic place can dominate the DPOAE measured in the stimulus at frequency fdp on the basilar membrane at the fdp canal, but normally only when the f2 / f1 ratio is smaller characteristic place. A probe tone similar in frequency to fdp ͑Shera and Guinan, 1999; Knight and Kemp, 2000, 2001; but with a more favorable forward transmission pathway ͒ Dhar et al., 2005 . In this experiment, the 2f1-f2 combination could have a higher level in the canal, but still be similar tone is therefore presumably initiated where the stimuli in- level to the combination tone near the fdp characteristic teract at the f2 characteristic place. Some energy at frequency place. Thus the probe and combination tones could possibly fdp then propagates basally to cross the middle ear and be beat together near the fdp characteristic place. Zwicker and recorded in the ear canal as the DPOAE. Some energy also Harris ͑1990͒ suggest cochlear origins of the gross discrep- propagates apically to the 2f1-f2 characteristic place where ancy between perceptual and DPOAE levels that include the basilar membrane responds as it would to an externally such transmission characteristics. Reverse transmission can supplied tone. affect emission level in a complex fashion ͑Shera and Zweig, The participants in the present study could reliably esti- 1992a, b͒, but Keefe ͑2002͒ suggests that near 500 Hz the mate the level of the combination tone using best-beats. The forward and reverse transfer functions are similar. Zhang and average best-beat probe tone level ͑45 dB SPL͒ was 20 dB Abbas ͑1997͒ also found a similar effect of middle ear pres-

J. Acoust. Soc. Am., Vol. 122, No. 2, August 2007 Purcell et al.: Magnetoencephalographic correlates of combination tones 999 sure on forward and reverse transmission. Thus it is possible lower in frequency than 2f1-f2. Suppression is most effective that at the fdp characteristic place in the cochlea, the probe either near the stimulus tones f2 and f1, or about 25 Hz tone caused a much larger response on the basilar membrane above the DPOAE frequency ͑Heitmann et al., 1998; Gaskill than the 2f1-f2 combination tone. and Brown, 1996; Talmadge et al., 1999; Konrad-Martin et If this is true, then the perceived beating may not have al., 2001͒. In the latter case, suppression is only present if a originated in the cochlea. Nonlinearities also occur at all significant 2f1-f2 place source is contributing to the DPOAE stages of neural processing above the cochlea. The probe measured in the ear canal. In the experiment reported here, tone would be transduced into a neural signal at its charac- the measured DPOAE should be from the stimulus region, teristic place on the basilar membrane near fdp. The 2f1-f2 and thus no suppression would be expected. Although there combination tone initiated near the f2 characteristic place is was no significant difference in average DPOAE level with present both in neurons servicing the f2 place, and in neurons the added probe tones, there were individuals whose re- ͑ ͒ servicing its own characteristic fdp place Kim et al., 1980 . sponse was slightly lower with the probe tones present. If the basilar membrane response at frequency fdp is small as suggested by the DPOAE, then beating between the large probe tone and small combination tone may occur in the B. MEG response auditory nervous system beyond the basilar membrane. The Dipoles were fit in each hemisphere using the 70-dB nonlinearities in these interactions may be such that a high SPL ECD stimulus. The average position values in Table I level probe beats best with a 2f1-f2 combination tone that can be compared to those reported previously using a similar was small in the cochlea but larger centrally. That combina- head model ͑Ross et al., 2000͒. The Ross et al. ͑2000͒ values tion tone may be larger in neural terms because of processes are all within 1.3 s.d. of the means found here ͑using the that enhance the detection of envelope frequencies from the present s.d. values͒. Since the ECD stimulus frequencies firing patterns of neurons servicing the regions of the f2 and were identical to those of the probe and combination tones, it f1 characteristic places. Beating could then occur in the cen- was assumed that the 40-Hz sources would be the same. tral nervous system between the 2f1-f2 combination tone and Sources are known to be tonotopic with carrier frequency the probe tone, which itself was mediated through neurons ͑Pantev et al., 1996͒, but it is not anticipated that they would servicing the fdp region of the cochlea. vary in location with stimulus level. A 40-Hz beat can be generated in the auditory nervous A 40-Hz beat was chosen to increase the likelihood of system through binaural presentation of stimuli that differ by detecting a response in the MEG signal due to the 40-Hz 40 Hz ͑Schwarz and Taylor, 2005, Draganova et al., unpub- peak in the transfer function of the steady-state response lished͒. The binaural auditory steady-state response has dif- ͑Ross et al., 2000͒. This required lowering the probe tone ferent characteristics from the response evoked by two tones frequency to 460 Hz from the 495.5 Hz that was used in the in the same ear. These findings clearly indicate that nonlin- psychoacoustic measurement of best-beats. The transfer earities of the auditory nervous system could generate a beat- function of the middle ear and cochlea should not change ing response distinct from that generated by nonlinearities in enough over this frequency range to affect the depth of the cochlea. modulation achieved between the probe and 2f1-f2 combina- Correlation analysis, however, showed no linear relation tone. tionship between DPOAE magnitude and either the psycho- Detection of the cortical correlate of the 2f1-f2 combi- acoustic or MEG measurements. As discussed earlier, this nation tone ͑evoked at perceptually relevant frequencies͒ was may be due to the unique processes involved in the genera- only possible indirectly because the level of the MEG tion and transmission of DPOAEs, or it may be because the steady-state response drops precipitously as the modulation beat response is generated above the cochlea in the auditory frequency approaches 100 Hz ͑Ross et al., 2000͒. In addi- nervous system. There was also no correlation between best- tion, detection is influenced by the practical limitations of beat probe tone levels and the MEG source space projection recording MEG signals from deep sources. Evidence sug- dipole moment magnitudes. Although these might be ex- gests that the envelope following response above 100 Hz is pected to be more closely related ͑both being mediated by mostly from sources in the brainstem ͑Herdman et al., 2002; the brain rather than the cochlea͒, one involves the complexi- Purcell et al., 2004͒. The indirect method used here was to ͑ ͒ ties of perception of beating near 5 Hz , and the other the cause beating between the 2f1-f2 combination tone and an synchronous firing of large populations of neurons at 40 Hz. external probe tone in Experiment 1, or between two differ- There may be large between-subject variations in the number ent 2f1-f2 combination tones in Experiment 2. Beating or of neurons that fire synchronously in response to amplitude envelope modulation evokes the envelope following re- modulation. Within-subjects, the changes in neural popula- sponse at the difference frequency between the tones acting tions and synchrony between 5 and 40 Hz may also be large. as stimuli. The quadratic combination tone has also been The net result of these complexities was that no correlations well studied psychoacoustically ͑e.g., Plomp, 1965; Gold- were found between the psychoacoustic and MEG correlates stein, 1966; Humes, 1979; 1980, 1985͒, and has been re- of the 2f1-f2 combination tone, or between the DPOAE mea- corded objectively along with the 2f1-f2 combination tone in surements and either the psychoacoustic or MEG measure- some physiological studies ͑e.g., Smoorenburg et al., 1976; ments. Kim et al., 1980; Gibian and Kim, 1982; Cheatham and Dal- No significant change in DPOAE level was expected los, 1997͒. The two have been employed together in psy- ͑ ͒ ͑ and none was found with added probe tones since they are choacoustics to facilitate judgments involving 2f1-f2 e.g.,

1000 J. Acoust. Soc. Am., Vol. 122, No. 2, August 2007 Purcell et al.: Magnetoencephalographic correlates of combination tones Goldstein, 1966͒. Combination tones have also been ob- 0.84 and 1.89 nA m in the left and right hemispheres, respec- served to evoke other combination tones ͑e.g., Goldstein et tively. The corresponding CT beat had source space al., 1978; Cheatham and Dallos, 1997͒. The purposeful use projection dipole moment magnitudes that were −18.0 and of both orders of combination tones here permitted a signal −17.7 dB smaller. These values are only about −3 dB derived from the 2f1-f2 combination tone to be measured in smaller than those between the ECD and probe beating cortex that would have been otherwise unavailable. stimuli discussed above for Experiment 1. Given the saturat- In Table II, the average test trial source space projection ing monotonic input/output relationship reported by Ross et dipole moment magnitude is larger ͑although not statistically al. ͑2000; their Fig. 7A͒, the combination tones in Experi- significantly so͒ in the right hemisphere contralateral to the ment 2 may have been slightly smaller than the average ones stimulus presentation to the left ear as expected from the in Experiment 1, or the achieved depth of modulation may results of Ross et al. ͑2005͒. The average test trial magni- have been slightly less favorable. tudes are smaller than those given in Table I for the dipole moments elicited with the ECD stimulus. The ECD stimulus V. CONCLUDING REMARKS level was 70 dB SPL, whereas the average best-beat probe These recordings have clearly demonstrated that the hu- tone level was close to 45 dB SPL. For a 250-Hz carrier that man auditory cortex responds to the 2f -f combination tone. ͑ ͒ 1 2 was amplitude modulated at 39 Hz, Ross et al. 2000 re- There were no significant correlations between the perceived ported that the dipole moment decreased to about 50%for a level of the combination tone and the magnitude of either the decrease in stimulus level of 25 dB. In the present study, the cortical MEG response or the cochlear DPOAE. The per- beating stimulus elicited a response that was less than 20% ceived combination tone is therefore likely mediated by pro- of the response to the ECD stimulus. While this could in cesses above those that produce the measured physiological ͑ small part be due to the presence of multiple tone pairs John responses. The MEG responses show that information about ͒ et al., 1998 , this increased difference than what might be the combination tone is transmitted to cortex, but later com- expected from the probe tone level is further evidence of the plex interactions between cortical areas likely mediate its disconnection between best-beat level and MEG magnitude. perception. If the Ross et al. ͑2000͒ data are representative of the situa- tion here, then the probe beating either did not have a modu- ACKNOWLEDGMENTS lation depth near 100%, or was lower in level than indicated by the probe tone level. Psychoacoustic methods involving Research supported by the Canadian Institutes of Health simultaneous probe and stimulus tones may overestimate the Research, and the National Institutes of Health Research. level of combination tones ͑Smooren- ͑ ͒ ͑ ͒ burg, 1972b;Shannon and Houtgast, 1980͒. However, it is Abdala, C. 1996 . “Distortion product otoacoustic emission 2f1-f2 amplitude as a function of f2/f1 frequency ratio and primary tone level separa- not clear where in the system the beat is generated and what tion in human adults and neonates,” J. Acoust. Soc. Am. 100, 3726–3740. the relative levels of the probe tone and combination tone Brown, D. K., Bowman, D. M., and Kimberley, B. P. ͑2000͒. “The effects of may be at that generator. maturation and stimulus parameters on the optimal f2/f1 ratio of the 2f1-f2 distortion product otoacoustic emission in neonates,” Hear. Res. 145,17– In Experiment 2, a successful attempt was made to elicit 24. a 40-Hz MEG response using two 2f1-f2 combination tones. Buunen, T. J., and Rhode, W. S. ͑1978͒. “Responses of fibers in the cat’s The purpose of this demonstration was to show that corre- auditory nerve to the cubic difference tone,” J. Acoust. Soc. Am. 64, lates of the 2f -f combination tone could be present in cor- 772–781. 1 2 Cheatham, M. A., and Dallos, P. ͑1997͒. “Intermodulation components in tex without the administration of an additional tone. Only inner hair cell and organ of corti responses,” J. Acoust. Soc. Am. 102, one of two participants in Experiment 2 showed a significant 1038–1048. 40-Hz MEG response. This was not unexpected since a re- Dhar, S., Long, G. R., Talmadge, C. L., and Tubis, A. ͑2005͒. “The effect of sponse would only be present if the two 2f -f combination stimulus-frequency ratio on distortion product otoacoustic emission com- 1 2 ponents,” J. Acoust. Soc. Am. 117, 3766–3776. tones were fortuitously of similar level. The two stimulus Dolphin, W. F. ͑1997͒. “The envelope following response to multiple tone tone pair levels were the same, but f2 / f1 was quite different pair stimuli,” Hear. Res. 110, 1–14. for the two pairs. Stimulus frequency ratio is known to affect Dolphin, W. F., and Mountain, D. C. ͑1993͒. “The envelope following re- ͑ sponse ͑EFR͒ in the mongolian gerbil to sinusoidally amplitude-modulated the level of DPOAEs e.g., Harris et al., 1989; Gorga et al., signals in the presence of simultaneously gated pure tones,” J. Acoust. 1993; Vinck et al., 1996; Abdala, 1996͒, and psychoacoustic Soc. Am. 94, 3215–3226. estimates ͑e.g., Goldstein, 1966; Smooren- Fastl, H. ͑1977͒. “Roughness and temporal masking patterns of sinusoidally burg, 1972a, b; Wilson, 1980͒. If the two 2f -f combination amplitude modulated broadband noise,” in Psychophysics and Physiology 1 2 of Hearing, edited by E. F. Evans and J. P. Wilson ͑Academic, London͒, tones were sufficiently different in level where they inter- pp. 403–414. acted, the depth of the envelope modulation would be insuf- Fastl, H. ͑1990͒. “The hearing sensation roughness and neuronal responses ficient to elicit a recordable 40-Hz MEG response. If the to AM-tones,” Hear. Res. 46, 293–295. ͑ ͒ absolute level of the combination tones were also too low, Furst, M., Rabinowitz, W. M., and Zurek, P. M. 1988 . “Ear canal acoustic distortion at 2f1-f2 from human ears: Relation to other emissions and there would also be no response. The fact that a very definite perceived combination tones,” J. Acoust. Soc. Am. 84, 215–221. 40-Hz response was elicited in one participant is evidence Gaskill, S. A., and Brown, A. M. ͑1990͒. “The behavior of the acoustic for the existence of both combination tones and their inter- distortion product, 2f1-f2, from the human ear and its relation to auditory action in the auditory system. sensitivity,” J. Acoust. Soc. Am. 88, 821–839. Gaskill, S. A., and Brown, A. M. ͑1996͒. “Suppression of human acoustic In response to the ECD stimulus, the single individual distortion product: Dual origin of 2f1-f2,” J. Acoust. Soc. Am. 100,3268– with significant responses in Experiment 2 had dipoles of 3274.

J. Acoust. Soc. Am., Vol. 122, No. 2, August 2007 Purcell et al.: Magnetoencephalographic correlates of combination tones 1001 Gibian, G. L., and Kim, D. O. ͑1982͒. “Cochlear microphonic evidence for Mäkelä, J. P., and Hari, R. ͑1987͒. “Evidence for cortical origin of the 40 Hz mechanical propagation of distortion products ͑f2-f1͒ and ͑2f1-f2͒,” Hear. auditory evoked response in man,” Electroencephalogr. Clin. Neuro- Res. 6, 35–59. physiol. 66, 539–546. Goldstein, J. L. ͑1966͒. “Auditory nonlinearity,” J. Acoust. Soc. Am. 41, Nuttall, A. L., and Dolan, D. F. ͑1990͒. “Inner hair cell responses to the 676–689. 2f1-f2 intermodulation distortion product,” J. Acoust. Soc. Am. 87, 782– Goldstein, J. L. ͑1969͒. “Aural combination tones,” in Frequency Analysis 790. and Periodicity Detection in Hearing, edited by R. Plomp and G. F. Pandya, P. K., and Krishnan, A. ͑2004͒. “Human frequency-following re- Smoorenburg ͑Sijthoff, Leiden, The Netherlands͒, pp. 230–247. sponse correlates of the distortion product at 2f1-f2,” J. Am. Acad. Audiol. Goldstein, J. L., Buchsbaum, G., and Furst, M. ͑1978͒. “Compatibility be- 15, 184–197. tween psychophysical and physiological measurements of aural combina- Pantev, C., Roberts, L. E., Elbert, T., Ross, B., and Wienbruch, C. ͑1996͒. tion tones,” J. Acoust. Soc. Am. 63, 474–485. “Tonotopic organization of the sources of human auditory steady-state Goldstein, J. L., and Kiang, N. Y. S. ͑1968͒. “Neural correlates of the aural responses,” Hear. Res. 101, 62–74. combination tone 2f1-f2,” Proc. IEEE 56, 981–992. Plomp, R. ͑1965͒. “Detectability threshold for combination tones,” J. Gorga, M. P., Neely, S. T., Bergman, B. M., Beauchaine, K. L., Kaminski, J. Acoust. Soc. Am. 37, 1110–1123. R., Peters, J., Schulte, L., and Jesteadt, W. ͑1993͒. “A comparison of Probst R., Loonsbury-Martin, B. L., and Martin, G. K. ͑1991͒. “A review of transient-evoked and distortion product otoacoustic emissions in normal- otoacoustic emissions,” J. Acoust. Soc. Am. 89, 2027–2067. hearing and hearing-impaired subjects,” J. Acoust. Soc. Am. 94, 2639– Purcell, D. W., John, S. M., Schneider, B. A., and Picton, T. W. ͑2004͒. 2648. “Human temporal auditory acuity as assessed by envelope following re- Gutschalk, A., Mase, R., Roth, R., Ille, N., Rupp, A., Hahnel, S., Picton, T. sponses,” J. Acoust. Soc. Am. 116, 3581–3593. ͑ ͒ W., and Scherg, M. 1999 . “Deconvolution of 40 Hz steady-state fields Rickman, M. D., Chertoff, M. E., and Hecox, K. E. ͑1991͒. “Electrophysi- reveals two overlapping source activities of the human auditory cortex,” ological evidence of nonlinear distortion products to two-tone stimuli,” J. Clin. Neurophysiol. 110, 856–868. Acoust. Soc. Am. 89, 2818–2826. ͑ ͒ Hall, J. L. 1975 . “Nonmonotonic behavior of distortion product 2f1-f2: Robles, L., Ruggero, M. A., and Rich, N. C. ͑1991͒. “Two-tone distortion in Psychophysical observations,” J. Acoust. Soc. Am. 58, 1046–1050. the basilar membrane of the cochlea,” Nature ͑London͒ 349, 413–414. Harris, F. P., Lonsbury-Martin, B. L., Stagner, B. B., Coats, A. C., and Robles, L., Ruggero, M. A., and Rich, N. C. ͑1997͒. “Two-tone distortion on ͑ ͒ Martin, G. K. 1989 . “Acoustic distortion products in humans: Systematic the basilar membrane of the chinchilla cochlea,” J. Neurophysiol. 77, 85 changes in amplitudes as a function of f2/f1 ratio,” J. Acoust. Soc. Am. , 2385–2399. 220–229. Ross, B., Borgmann, C., Draganova, R., Roberts, L. E., and Pantev, C. Heitmann, J., Waldmann, B., Schnitzler, H. U., Plinkert, P. K., and Zenner, ͑2000͒. “A high-precision magnetoencephalographic study of human au- H. P. ͑1998͒. “Suppression of distortion product otoacoustic emissions ditory steady-state responses to amplitude-modulated tones,” J. Acoust. ͑DPOAE͒ near 2f1-f2 removes DP-gram fine structure-Evidence for a sec- Soc. Am. 108, 679–691. ondary generator,” J. Acoust. Soc. Am. 103, 1527–1531. Ross, B., Herdman, A. T., and Pantev, C. ͑2005͒. “Right hemispheric later- Herdman, A. T., Lins, O., Van Roon, P., Stapells, D. R., Scherg, M., and ality of human 40 Hz auditory steady-state responses,” Cereb. Cortex 15, Picton, T. W. ͑2002͒. “Intracerebral sources of human auditory steady- 2029–2039. state responses,” Brain Topogr 15, 69–86. Ross, B., Picton, T. W., and Pantev, C. ͑2002͒. “Temporal integration in the Humes, L. E. ͑1979͒. “Perception of the simple difference tone ͑f2-f1͒,” J. Acoust. Soc. Am. 66, 1064–1074. human auditory cortex as represented by the development of the steady- Humes, L. E. ͑1980͒.“GrowthofL͑f2-f1͒ and L͑2f1-f2͒ with input level: state magnetic field,” Hear. Res. 165, 68–84. ͑ ͒ Influence of f2/f1,” Hear. Res. 2, 115–122. Schoonhoven, R., Boden, C. J., Verbunt, J. P., and de Munck, J. C. 2003 . Humes, L. E. ͑1985͒. “Cancellation level and phase of the ͑f2-f1͒ distortion “A whole head MEG study of the amplitude-modulation-following re- product,” J. Acoust. Soc. Am. 78, 1245–1251. sponse: Phase coherence, group delay and dipole source analysis,” Clin. John, M. S., Lins, O. G., Boucher, B. L., and Picton, T. W. ͑1998͒. “Multiple Neurophysiol. 114, 2096–2106. ͑ ͒ auditory steady state responses ͑MASTER͒: Stimulus and recording pa- Schwarz, D. W., and Taylor, P. 2005 . “Human auditory steady state re- rameters,” Audiology 37, 59–82. sponses to binaural and monaural beats,” Clin. Neurophysiol. 116,658– John, M. S., and Picton, T. W. ͑2000͒. “MASTER: A windows program for 668. ͑ ͒ recording multiple auditory steady-state responses,” Comput. Methods Shannon, R. V., and Houtgast, T. 1980 . “Psychophysical measurements Programs Biomed. 61, 125–150. relating suppression and combination tones,” J. Acoust. Soc. Am. 68,825– Jones, A. T. ͑1935͒. “The discovery of difference tones,” Am. Phys. Teach. 829. ͑ ͒ 3, 49–51. Shera, C. A., and Guinan, J. J., Jr. 1999 . “Evoked otoacoustic emissions Keefe, D. H. ͑2002͒. “Spectral shapes of forward and reverse transfer func- arise by two fundamentally different mechanisms: A taxonomy for mam- tions between ear canal and cochlea estimated using DPOAE input/output malian OAEs,” J. Acoust. Soc. Am. 105, 782–798. functions,” J. Acoust. Soc. Am. 111, 249–260. Shera, C. A., and Zweig, G. ͑1992a͒. “Middle-ear phenomenology: The Kemp, D. T., and Brown, A. M. ͑1984͒. “Ear canal acoustic and round view from the three windows,” J. Acoust. Soc. Am. 92, 1356–1370. window electrical correlates of 2f1-f2 distortion generated in the cochlea,” Shera, C. A., and Zweig, G. ͑1992b͒. “Analyzing reverse middle-ear trans- Hear. Res. 13, 39–46. mission: Noninvasive gedankenexperiments,” J. Acoust. Soc. Am. 92, Kim, D. O., Molnar, C. E., and Matthews, J. W. ͑1980͒. “Cochlear mechan- 1371–1381. ics: Nonlinear behavior in two-tone responses as reflected in cochlear- Siegel, J. H. ͑1994͒. “Ear-canal standing waves and high-frequency sound nerve-fiber responses and in ear-canal sound pressure,” J. Acoust. Soc. calibration using otoacoustic emission probes,” J. Acoust. Soc. Am. 95, Am. 67, 1704–1721. 2589–2597. Knight, R. D., and Kemp, D. T. ͑2000͒. “Indications of different distortion Smoorenburg, G. F. ͑1972a͒. “Audibility region of combination tones,” J. product otoacoustic emission mechanisms from a detailed f1, f2 area Acoust. Soc. Am. 52, 603–614. study,” J. Acoust. Soc. Am. 107, 457–473. Smoorenburg, G. F. ͑1972b͒. “Combination tones and their origin,” J. Knight, R. D., and Kemp, D. T. ͑2001͒. “Wave and place fixed DPOAE Acoust. Soc. Am. 52, 615–632. maps of the human ear,” J. Acoust. Soc. Am. 109, 1513–1525. Smoorenburg, G. F., Gibson, M. M., Kitzes, L. M., Rose, J. E., and Hind, J. Kohlrausch, A., Fassel, R., and Dau, T. ͑2000͒. “The influence of carrier E. ͑1976͒. “Correlates of combination tones observed in the response of level and frequency on modulation and beat-detection thresholds for sinu- neurons in the anteroventral cochlear nucleus of the cat,” J. Acoust. Soc. soidal carriers,” J. Acoust. Soc. Am. 108, 723–734. Am. 59, 945–962. Konrad-Martin, D., Neely, S. T., Keefe, D. H., Dorn, P. A., and Gorga, M. P. Strickland, E. A. ͑2000͒. “The effects of frequency region and level on the ͑2001͒. “Sources of distortion product otoacoustic emissions revealed by temporal modulation transfer function,” J. Acoust. Soc. Am. 107, 942– suppression experiments and inverse fast Fourier transforms in normal 952. ears,” J. Acoust. Soc. Am. 109, 2862–2879. Talmadge, C. L., Long, G. R., Tubis, A., and Dhar, S. ͑1999͒. “Experimental Lins, O. G., Picton, P. E., Picton, T. W., Champagne, S. C., and Durieux- confirmation of the two-source interference model for the fine structure of Smith, A. ͑1995͒. “Auditory steady-state responses to tones amplitude- distortion product otoacoustic emissions,” J. Acoust. Soc. Am. 105, 275– modulated at 80–110 Hz,” J. Acoust. Soc. Am. 97, 3051–3063. 292.

1002 J. Acoust. Soc. Am., Vol. 122, No. 2, August 2007 Purcell et al.: Magnetoencephalographic correlates of combination tones Vinck, B. M., De Vel, E., Xu, Z. M., and Van Cauwenberge, P. B. ͑1996͒. Yost, W. A. ͑2000͒. Fundamentals of Hearing: An Introduction, 4th ed. “Distortion product otoacoustic emissions: A normative study,” Audiology ͑Elsevier Academic, San Diego͒. 35, 231–245. Zhang, M., and Abbas, P. J. ͑1997͒. “Effects of middle ear pressure on von Helmholtz, H. L. F. ͑1877/1954͒. On the Sensations of Tone ͓translation otoacoustic emission measures,” J. Acoust. Soc. Am. 102, 1032–1037. ͑ ͒ of Die Lehre von den Tonempﬁndugen͔͑Dover, New York͒. Zwicker, E. 1955 . “Der ungewöhnliche amplitudengang der nichtlinearen 5 Whitehead, M. L., Stagner, B. B., Lonsbury-Martin, B. L., and Martin, G. K. verzerrungen des ohres,” Acustica , 67–74. Zwicker, E. ͑1981͒. “Dependence of level and phase of the ͑2f1-f2͒- ͑1995͒. “Effects of ear-canal standing waves on measurements of cancellation tone on frequency-range, frequency difference, level of pri- 98 distortion-product otoacoustic emissions,” J. Acoust. Soc. Am. , 3200– maries, and subject,” J. Acoust. Soc. Am. 70, 1277–1288. 3214. Zwicker, E., and Harris, F. P. ͑1990͒. “Psychoacoustical and ear canal can- ͑ ͒ Wilson, J. P. 1980 . “The combination tone, 2f1-f2, in psychophysics and cellation of ͑2f1-f2͒-distortion products,” J. Acoust. Soc. Am. 87, 2583– ear-canal recording,” in Psychophysical, Physiological and Behavioral 2591. Studies in Hearing, edited by G. Van der Brink and F. A. Bilsen ͑Delft Zwislocki, J. ͑1953͒. “Acoustic attenuation between the ears,” J. Acoust. University, Delft, The Netherlands͒, pp. 43–50. Soc. Am. 25, 752–759.

J. Acoust. Soc. Am., Vol. 122, No. 2, August 2007 Purcell et al.: Magnetoencephalographic correlates of combination tones 1003